Simulation engines: video games in the age of dual technologies

It would be a mis­take to reduce video games to only their visu­als, their nar­rat­ives or their tar­get mar­ket. But their tech­nic­al power lies else­where: they cre­ate envir­on­ments in which action takes place. A game engine does not merely dis­play a scene; it organ­ises a space gov­erned by con­straints with­in which play­ers can act, fail, try again, optim­ise their actions and reveal dis­crep­an­cies between sur­face rules and under­ly­ing rules.

A sim­u­la­tion engine imposes a prac­tic­al frame­work: act­ors, pos­i­tions, states, trans­itions, vis­ible rules, under­ly­ing rules, thresholds and feed­back. A use­ful sim­u­la­tion is not the one that most closely resembles real­ity. It is the one that effect­ively addresses a prob­lem. It must sim­pli­fy real­ity enough to make it man­age­able, but not to the point that the les­sons learned are misleading.

The Sig­nal Scope sim­u­la­tion, developed for the École Nationale Supérieure de la Police, and more recently Oper­a­tion Acon­it, con­duc­ted on behalf of the French Army, illus­trate this approach in an oper­a­tion­al con­text. Their aim is not to add a fun ele­ment to an exist­ing train­ing pro­gramme, but to cre­ate a con­trolled envir­on­ment for address­ing a spe­cif­ic prob­lem. Par­ti­cipants work with­in a situ­ation struc­tured by roles, imper­fect inform­a­tion, expli­cit con­straints, room for man­oeuvre and delayed effects. Decisions leave a trail; assess­ments can be ques­tioned; bound­ar­ies emerge with­in the group dynam­ic; and the debrief­ing then com­pares the paths taken, the dead ends and the work­arounds. Acon­it reveals what a simple read­ing often hides: the way in which a group alloc­ates its atten­tion, pri­or­it­ises risks, accepts uncer­tainty and trans­lates an instruc­tion into actu­al beha­viour. The value of Acon­it there­fore does not lie in the game itself. It lies in the pro­duc­tion of action­able oper­a­tion­al mater­i­al: what was observed, decided, lost, cir­cum­ven­ted or mis­in­ter­preted. The game becomes a frame­work of con­straints, obser­va­tion and decision-making.

Meth­ods such as tech­nic­al scen­ario-build­ing, design fic­tion and sci­ence-fic­tion pro­to­typ­ing then come into play. Their value does not lie in the nar­rat­ive. It lies in their abil­ity to decon­struct rep­res­ent­a­tions, to make lat­ent con­sequences vis­ible, to allow for objec­tions that are dif­fi­cult to voice with­in a hier­arch­ic­al organ­isa­tion, and to test the lim­its of a sys­tem.  Mul­tiver­sal­ism must be under­stood in this prac­tic­al sense. It is not about cel­eb­rat­ing spec­tac­u­lar vir­tu­al worlds. It is about gen­er­at­ing mul­tiple branches of the same prob­lem: mul­tiple scen­ari­os, mul­tiple sets of con­straints, mul­tiple reac­tions from act­ors, mul­tiple crit­ic­al thresholds. The mul­ti­verse becomes a meth­od for com­par­ing worlds of action. It allows us to return to real­ity with bet­ter hypotheses.

Vin­cent Bon­tems. Yes, provided we con­sider sev­er­al dif­fer­ent per­spect­ives. For the pro­gram­mer, the engine com­prises the com­pon­ents that cal­cu­late the geo­metry and pseudo-phys­ics of the envir­on­ment. For the play­er, it becomes the avatar’s asso­ci­ated envir­on­ment, with resources and con­straints that give rise to phys­ic­al and intel­lec­tu­al tech­niques. For the a ‘tech­no­logy philo­soph­er’, the digit­al object is estab­lished with­in the cir­cuit link­ing the involved sub­ject, the machine, the images and the actions. This cir­cuit remains asym­met­ric­al: the play­er does not, strictly speak­ing, modi­fy the engine. We must there­fore ana­lyse what this engine encloses, opens up and stabilises.

It enables us to look bey­ond the sur­face of these uses. We must dis­tin­guish between what changes, such as images, sounds, scripts and object­ives, and what remains con­stant, such as cer­tain cal­cu­la­tions, rhythms, move­ment con­straints, forms of anti­cip­a­tion or learn­ing pro­cesses. Mod­ding prac­tices clearly illus­trate this cre­at­ive sta­bil­ity: we trans­form the exper­i­ence, but we work with­in the engine’s exist­ing frame­work. A com­par­at­ive study of engines thus enables us to cat­egor­ise learn­ing pro­cesses. It would, how­ever, be a mis­take to believe that sim­u­lated learn­ing is identic­al to real-world exper­i­ence. There is always a gap, even in a flight simulator.

They can be cat­egor­ised along three axes: sim­u­la­tion, that is, the real­ism of entit­ies’ beha­viour; immer­sion, that is, sens­ory, aes­thet­ic and nar­rat­ive engage­ment; and gami­fic­a­tion, that is, the mobil­isa­tion of emo­tion­al response through a reward sys­tem. Mar­ket dis­tinc­tions mat­ter less than the com­bin­a­tion of these three dimen­sions. In a pro­ject for the armed forces on wound man­age­ment, the aim was not to make the wound pleas­ant or real­ist­ic for its own sake, but to make its con­sequences man­age­able with­in an enga­ging exper­i­ence. The key cri­terion was not the label of the sys­tem, but the exper­i­ence to be created.

Mul­tiver­sal­ism involves ima­gin­ing oth­er worlds to think about the world dif­fer­ently. It is based on the inter­play between the “what if?” of the ima­gin­a­tion and the “yes, but…” of crit­ic­al ration­al­ity. Any scen­ario-build­ing reduces a situ­ation to a few factors, the vari­ations of which are then com­bined. The chal­lenge lies in choos­ing the right vari­ables and under­stand­ing why we vary them. Explor­ing pos­sib­il­it­ies encour­ages a shift away from fixed per­spect­ives; focus­ing on a tech­no­lo­gic­al tra­ject­ory, its depend­en­cies and its tip­ping points, on the oth­er hand, falls more with­in the remit of Net Tech­no­lo­gic­al Assess­ment. The two approaches should be integrated.

We must doc­u­ment and sim­u­late accord­ing to thresholds. The approach you pro­posed dur­ing the debrief­ing of the work­shop on anti­cip­at­ing crit­ic­al thresholds leads to this: determ­in­ing, in the present, the vari­ables that cause a situ­ation to tip over, then test­ing them accord­ing to the rel­ev­ant scales. Vir­tu­al sim­u­la­tion can help to isol­ate these vari­ables, gen­er­ate new observ­ables and test strategies for resi­li­ence. But it does not replace doc­u­ment­a­tion. It organ­ises it around thresholds, link­ages between tech­no­lo­gies and asso­ci­ated envir­on­ments, and scales of observation.

It becomes a meth­od when it pro­duces three effects. Firstly, decon­tex­tu­al­isa­tion: it strips away the cer­tain­ties of the present. Secondly, amp­li­fic­a­tion: it pushes the con­sequences bey­ond imme­di­ately access­ible scales. Finally, the “umbrella” effect: it allows dis­sent­ing view­points to be expressed with­in a hier­arch­ic­al organ­isa­tion. Say­ing “I wouldn’t want to live in this world” may be more accept­able than say­ing “no” out­right. For every prob­lem, there is a form of pro­to­typ­ing: one can sci­en­ti­fi­cise a work of fic­tion, recon­struct a tech­no­logy with­in a con­strained uni­verse, or ima­gine an altern­at­ive tra­ject­ory to test an exist­ing system.

We need to look at the issue from the oppos­ite angle. Today’s dual-use tech­no­lo­gies are integ­rated from the out­set into the social, eco­nom­ic and polit­ic­al spheres, but their civil­ian uses are insep­ar­able from their mil­it­ary uses. Robot­isa­tion, the pro­lif­er­a­tion of drones and arti­fi­cial intel­li­gence are already chan­ging the paradigms of con­flict and cog­nit­ive war­fare. Meth­ods for assess­ing the matur­ity of tech­nic­al devel­op­ments remain use­ful, wheth­er they be TRIZ or Simondon’s genet­ic mechan­o­logy. But these must be sup­ple­men­ted by a form of sci­ence-fic­tion pro­to­typ­ing designed to gauge the pos­sible trans­form­a­tions in the bal­ance of power. There is no mir­acle meth­od. There are com­ple­ment­ary oper­a­tion­al methods.

The shift is simple: video games should not be impor­ted as a mod­el but stud­ied as an already estab­lished tech­nic­al labor­at­ory. They have estab­lished ways of cre­at­ing spaces, mak­ing avatars act, man­aging inter­ac­tions, script­ing con­straints and meas­ur­ing beha­viours. This frame­work is now being applied to train­ing, intern­al secur­ity, industry, defence, cyber­se­cur­ity, crisis man­age­ment and decision sup­port.

For dual-use tech­no­lo­gies, the assess­ment must there­fore address three prac­tic­al ques­tions. What does the solu­tion do? What world of action does it make pos­sible? Is this world tech­nic­ally, indus­tri­ally and oper­a­tion­ally sus­tain­able? Busi­ness intel­li­gence iden­ti­fies the act­ors, depend­en­cies and tra­ject­or­ies. Sim­u­la­tion tests scen­ari­os. Dis­cip­lined mul­tiver­sal­ism com­pares thresholds and pos­sible branches. Their integ­ra­tion trans­forms cre­ativ­ity into an assess­ment tool, and tech­no­logy watch into a decision-mak­ing method.